Figure 1a shows the RHEED pattern with sharp spots indicative of rough surface caused by surface diffusion of As atoms. High-resolution AFM image (Fig. 1b) reveals surface decomposition into oval-like plates surrounded by depressions seen in the dark contrast. The surface is densely covered with irregular structures with RMS roughness of 1.8 nm. X-ray photoemission spectroscopy (XPS) and AFM auxiliary experiments suggest that the decomposed surface is As rich. In addition, the plate density and the depressions size vary with temperature and annealing time. The decomposed surface proved to be essential where the plates act as templates for the Ag assembly, and the depressions provide free space (sinks) for the MWCNTs to accommodate.

Figure 2a shows a typical wide scan AFM image recorded in the final stages of dewetting of the toluene on the decomposed surface of GaAs. Clearly, toluene forms broken net-like patterns, and sometime hexagonal structures extending over the whole surface occupying the spacing between the plates. Interestingly, the calculated RMS roughness value is 0.2 nm less compared to that found without the toluene (i.e. decomposed surface shown in Fig. 1b). This reduction in the roughness is attributed to the smoothing effect induced by toluene as it fills the depressions.

Assembling of MWCNTs on the GaAs decomposed surface is shown in Fig. 2b. A MWCNTs net-like pattern similar to that of toluene can be clearly seen. The image shown was attained after desorption of toluene at 300°C for 30 min. Hence, the observed features are only attributed to the MWCNTs with average thickness of 8 nm and the decomposed surface. This is well reflected from the high-resolution 3D image shown in Fig. 2c.

At this point, we turn our discussion to the wetting of the decomposed surface by a colloidal system consisting of 0.1% concentration of Ag nanoparticles dissolved in toluene. The effect of Ag nanoparticles on the toluene wetting behavior and its hydrodynamic is huge. Figure 3a depicts this effect where the wetting by the Ag colloidal particles induces separated array of droplets. Similar droplets have been observed following the wetting of a toluene solution composed of polystyrene and Ag nanoparticles on silicon substrates 30. The droplets have an average diameter of 1 μm, average distance of 1.9 μm and droplet density of 2.5 × 10⁷ cm⁻². To investigate the self-assembly of the Ag nanoparticles within the microdroplet structures, the samples were annealed at 300°C. From Fig. 3b, we notice two types of aggregates formed after annealing: the normal plates of the decomposed GaAs surface and spot structures surrounded by rings (e.g. enclosed by square). Separate spots and empty rings have been reported by Nobuhiko et al. 30 while observing polystyrene and Ag nanoparticles system. To the best of our knowledge, this is the first report on a combination of Ag spot and Ag ring in one structure. Different mechanisms have been proposed for the ring and spot formation (sometimes refereed as aggregates or islands). These are as follows: capillary flow-induced localization along the contact line (the border between the liquid film and the substrate or the three-phase line of the dewetting solution film) 3031, capillary flow 32, kinetically driven islands formation 33 and pinning of dewetting hole 34. Here, we argue that the first three mechanisms are responsible at different stages for the combined structure formation as schematically illustrated in Fig. 4. Initially, we adapt the capillary flow-induced localization along the contact line only to explain the formation of droplets (Fig. 3a) before the evaporation. Firstly, the contact line moves toward the liquid phase and then the Ag nanoparticles are captured by capillary flow and collected at contact line forming a sinusoidal interface leading to the microdroplets formation (Fig. 4a, b, c). Once the evaporation starts, the droplets shrink due to the annealing of the sample. This is reinforced by the fact that the observed size of the droplet is four times larger compared to that of the combined structure as shown in Fig. 3c. At this stage, two simultaneous events are taking place as shown in Fig. 4d, e. These events are very fast to be experimentally detected by many techniques including AFM used in this work. Therefore, here we provide only theoretically grounded 3233 explanation without any direct evidence of their occurrence. The first event is the compensation by outward flow to restore the shrinkage (capillary flow mechanism) and deepening of the solvent–air interface of the contracted droplet. As the evaporation continues, the contact line recedes and lays its accumulated particles (kinetically driven islands formation mechanism as shown in Fig. 4d) on the substrate to form the spot structure, and the outward flow forms the ring surrounding the spot as indicated in Fig. 4f.

Next, we attempt to examine a complete system consisting of Ag and MWCNTs on the decomposed GaAs surface. Figure 5a shows the AFM image of such system after annealing of the sample at 300°C for 30 min. Spherical droplets and combined structures such as those found and shown in Fig. 3a, 3b are not any more observed. Instead, the film breaks into remarkable compact 2D Ag islands seen on the top of arsenic plate and surrounded by MWCNTs.

In order to quantify our observations and to investigate the effects of surface decomposition on the assembly of MWCNTs and Ag nanoparticles, PSD analysis was carried out. Figure 5b represents the PSD profiles of these systems. Both low and high spatial frequency regions of the PSD have been influenced by the different sample preparation conditions. However, in the high frequency region all systems obey the inverse power law (fractal model) indicating a strong presence of fractal components (i.e. the effect of As plates in all systems) and given by 35,

where f is the frequency, v and K are the spectral indices and spectral strength, respectively. In addition, the slope of the PSD at this region is almost the same indicating the same fractal spectral indices (v) value of 5.1. Since v is related to the roughness exponent (α) by the equation α = (v − d)/2 36, where the line scan dimension d is 2, α = (5.1–2)/2 = 1.5. The high frequency PSD analysis supports the idea that the plates have strong influence on the distribution of MWCNTs and toluene. As seen from the PSD, the low frequency parts vary and represent predominantly the aggregates. Both decomposed GaAs/Toluene and decomposed GaAs/MWCNTs system show ‘knee’ features (indicated by arrows) in PSD profiles at the interface between the two frequency regions. Their presence indicates aggregation of either toluene or MWCNTs at the depressions as observed in the AFM images shown in Fig. 2a, 2b. These features are absent in the decomposed GaAs surface. Apart from substrate’s fractals (plates), the aggregates in the depressions can be considered as fractals developed during the growth stage. In this case, these extra fractals can be described by the ABC model which is given by 37,

with A, B, C being parameters involved in describing the RMS roughness and the correlation length in wide frequency range. The PSD profile for the GaAs/Toluene system shown in the inset of Fig. 5b is indeed well fitted with this model.

Based on AFM experimental observations and PSD data analysis, a conceptual growth model of GaAs/MWCNTs/Ag system is proposed and schematically illustrated in Fig. 6. Once the substrate is etched, its surface decomposes into plates due to the thermal treatment. The density and the spacing between the plates depend on the decomposition temperature and time. Upon spin coating, the majority of MWCNTs dissolved in toluene occupying the spacing and act as boundaries to isolate the plates. If Ag nanoparticles are deposited directly on decomposed substrate without MWCNTs, combined structures (spots + rings) are formed. However, incorporation of MWCNTs cause the Ag to form compact assembly of Ag islands as observed in Fig. 5a and illustrated in Fig. 6d.

Figure 7 shows the PL spectra measured at room temperature for the different components in the GaAs/MWCNTs/Ag system. The intensity of the PL spectra shows both photoluminescence enhancement and quenching of the GaAs/MWCNTs/Ag system. The PL enhancement is obtained with the incorporation of MWCNTs of diameter 8 nm, whereas quenching is observed with the use of MWCNTs of diameter ~50 nm. Our results match with theoretical studies predicting that carbon nanotubes, used as spectroscopic enhancers, filled with Ag nanoparticles lead to an enhancement of optical response 29. The PL of GaAs/MWCNTs/Ag system can be tuned by varying the diameter of MWCNTs, though the dependency mechanism of the tube diameter with optical response is still unclear. From Fig. 7, the silver nanoparticles PL signal was found to be around 400 nm, and the addition of MWCNTs to the silver/GaAs system resulted in the intensity variation. It is also observed that the emission peak energy of the silver nanoparticles shifts from 3.12 eV (397 nm) to 3.2 eV (387 nm), i.e. inclusion of MWCNTs of higher diameter (≈50 nm) into GaAs/Ag system results in a blue shift in the emission peak energy with reduction in the intensity of the PL spectrum. It is worth mentioning at this stage that although light emission, resulting for surface plasmons, may have very short life time, its effect on the PL signal can be caught by the spectrometer when operating in appropriate conditions as detailed in the Sect. “Experimental Details”. The peaks indicated by arrows in Fig. 7 could possibly be attributed to the coupling between MWCNTs and Ag nanoparticles, similar to that reported for Au/MWCNTs system 38. Regardless of the silver MWCNTs coupling, PL peaks of high index decomposed GaAs matrix presented in the inset of Fig. 7 show a slight red-shift with reduction in the PL intensity of the well-known GaAs peak around 870 nm 29. This reduction is due to the fact that GaAs matrix is partially covered by either silver nanoparticles or MWCNTs. Hence, large electron diffusion length of surface deposited Ag on the GaAs surface causes a decrease in the number of electrons useful for recombination processes. The reason for red-shift needs more exploration to check whether quantum confinement effect of decreasing particle size is the only determining factor 39.